Large language models as collaborators in molecular thermodynamics: a case study of the LLM-assisted development of SAFT-VR IMP, an equation of state for polar fluids
Year: 2026
Type: preprint
Abstract: Whether large language models (LLMs) can contribute to non-trivial science, beyond summarising literature and writing routine code, is an open question. We report a case study in which an LLM (Claude) was asked to formulate, implement, verify and parameterise a new equation of state of the statistical associating fluid theory (SAFT) family. The target was a SAFT-VR variant for polar and dipolar molecules in which the segments interact through the isotropic multipolar potential (IMP) of Müller and Gelb, expressed as a temperature-dependent sum of Sutherland terms and treated with the Barker–Henderson perturbation theory of SAFT-VR Mie and SAFT-VR Sum. Over three sessions the LLM produced a stepby-step derivation, a 600-line implementation in the open-source Julia package Clapeyron.jl, a suite of 97 verification tests, a parameter-estimation pipeline applied to DIPPR data for six cyclic ethers, ketones and a sulfide, and a comparison with the group-contribution SAFT-γ
Source: ChemRxiv
Institutions: IFP Énergies nouvelles, Imperial College London
Language: English
Subfield: Physical and Theoretical Chemistry
Field: Chemistry
Domain: Physical Sciences
Open Access status: green
Funder: Agence Nationale de la Recherche
OpenAlex ID: W7220947328